AI Agents Could Be Crypto’s Next Breakthrough – If We Shift From Hype to Utility

Alec Goh is Head of HTX Ventures, the global investment arm of cryptocurrency exchange HTX. In this Op-Ed, Alec argues that the convergence of AI Agents and Web3 holds transformative potential, but requires the industry to avoid common crypto pitfalls.

Building Sustainable Value in Crypto + AI

I have navigated through multiple market cycles in the digital asset space—most were driven by narratives and authentic technological advancements. Only a few changed the game entirely.

In the last six months, it has become clear to me that the intersection of AI Agents and Web3 has the potential to encompass both elements, but only if the sector can move past early speculation and build for long-term value for end users.

We’ve seen this story before. In 2017, the ICO boom outpaced any meaningful product development. In 2021, NFTs exploded into the mainstream before many platforms had even built the underlying infrastructure for creators and collectors. In 2017, the ICO boom outpaced any meaningful product development. In both cases, the correction was as swift as the rise. And in both cases, the correction was as swift as the rise.

The recent explosion of AI Agent tokens followed a similar arc. Dozens of projects launched under the banner of “Crypto + AI”, attracting large inflows of capital and attention. Some achieved billion-dollar valuations within weeks. But as with past cycles, speculative capital alone does not create sustainable ecosystems. As investor enthusiasm cooled, so did user engagement—and most early token economies have struggled to retain relevance.

Yet amid that volatility, something changed.

In Q4 2024, Deepseek introduced a technical breakthrough that caught the attention of our team at HTX Ventures: a large-scale model trained entirely through reinforcement learning, with no reliance on human-labeled data. In practical terms, this dramatically reduces the cost of developing autonomous, intelligent agents. For the first time, it became feasible to deploy AI agents across Web3 with scalability and adaptability without the prohibitive infrastructure typically associated with AI development.

This shift matters. If crypto is to fulfill its promise as a decentralized, permissionless, and programmable financial layer, then agents capable of autonomous decision-making will become essential infrastructure. And yet, we are still some way from that vision.

What concerns me is that we are once again at risk of mistaking momentum for maturity. A handful of promising frameworks—like Eliza—are beginning to offer composable agent tooling across Ethereum, Solana, and TON. Launchpads such as Virtual and Clanker have helped accelerate token distribution. Application-level products like GRIFFAIN and NEUR are experimenting with AI-driven asset management and governance tools. But these remain early signals, not proven systems.

Too many AI Agent projects are still designed for speculation, not sustainability. And if that continues, we will face another cycle of capital flight, unmet expectations, and delayed progress.

That would be a missed opportunity.

Because the use cases are real. Defi strategies that are too complex for most users could be simplified through autonomous portfolio managers. Governance systems could be enhanced by agents that parse proposals, evaluate precedent, and vote based on user-defined values. And cross-chain automation could become seamless through agents acting on user intent.

In each of these cases, the need is not hypothetical—it is already here. What’s missing is consistent execution, user retention, and functional token economics.

So where do we go from here?

First, we must hold AI Agent projects to the same level of scrutiny we apply to DeFi and infrastructure protocols. That starts with moving beyond vague narratives to rigorous project frameworks—asking difficult but essential questions such as:

• Who are the users?
• How does it provide essential utility?
• Where is the revenue?
• How will the protocol remain relevant after token incentives end?

Protocols must demonstrate not only technical potential, but also real-world usage, sustainable revenue models, and measurable impact. This is the difference between a proof of concept and a protocol worth building on.

Second, on top of applications, the industry needs to provide a solid infrastructure that enables the success of AI Agents on-chain, this includes Agent frameworks, data layers, computational infrastructures and more. HTX Ventures will keep a close eye on these fundamental developments because some parts from this space will become as important as EVM compatibility became for L1s in 2020.

And finally, we need to foster a culture that prioritizes iteration and transparency. Not every project will get it right on the first attempt. But those that do will likely define the next frontier of Web3.

The convergence of AI and crypto is not a novelty. It is a necessary evolution. But if we are to seize this moment, the industry must focus less on capturing attention and more on delivering value.

This isn’t just a thesis for HTX Ventures. It’s a conviction grounded in past lessons and future potential.

We have the tools. We have the builders. What we need now is discipline.

Source